High-Throughput, Living Single-Cell, Multiple Secreted Biomarker Profiling Using Microfluidic Chip and Machine Learning for Tumor Cell Classification.

Abstract

Secreted proteins provide abundant functional information on living cells and can be used as important tumor diagnostic markers, of which profiling at the single-cell level is helpful for accurate tumor cell classification. Currently, achieving living single-cell multi-index, high-sensitivity, and quantitative secretion biomarker profiling remains a great challenge. Here, a high-throughput living single-cell multi-index secreted biomarker profiling platform is proposed, combined with machine learning, to achieve accurate tumor cell classification. A single-cell culture microfluidic chip with self-assembled graphene oxide quantum dots (GOQDs) enables high-activity single-cell culture, ensuring normal secretion of biomarkers and high-throughput single-cell separation, providing sufficient statistical data for machine learning. At the same time, the antibody barcode chip with self-assembled GOQDs performs multi-index, highly sensitive, and quantitative detection of secreted biomarkers, in which each cell culture chamber covers a whole barcode array. Importantly, by combining the K-means strategy with machine learning, thousands of single tumor cell secretion data are analyzed, enabling tumor cell classification with a recognition accuracy of 95.0%. In addition, further profiling of the grouping results reveals the unique secretion characteristics of subgroups. This work provides an intelligent platform for high-throughput living single-cell multiple secretion biomarker profiling, which has broad implications for cancer investigation and biomedical research.